Skip to Content
Applied SciencesApplied Sciences
  • Article
  • Open Access

24 February 2026

15 Pages

Road Accidents in the Context of Infrastructure and Economic Factors

and
1
Department of Transport, Stanislaw Staszic State University of Applied Sciences in Pila, Podchorazych 10 Street, 64-920 Pila, Poland
2
Department of Automobile Engineering, Faculty of Transport Engineering, Vilnius Gediminas Technical University, Plytinės g. 25, 10105 Vilnius, Lithuania
*
Author to whom correspondence should be addressed.
This article belongs to the Section Transportation and Future Mobility

Abstract

Every year, road accidents cause significant human and social losses, posing one of the key challenges for public policy in Poland. The aim of this article is to quantitatively assess the relationship between selected infrastructural and economic conditions and the scale of road accidents in Poland in the period 2010–2024. The analysis was carried out using a log–linear regression model, which allows the results to be interpreted in terms of elasticity. The dependent variable was the total number of road accidents, while the set of explanatory variables included the density of paved roads, the length of expressways and motorways, urban population density, the level of private car ownership, and average gross wages. The results indicate that the development of road infrastructure and an increase in the population’s income contribute to reducing the number of accidents, while the growing number of passenger cars significantly increases the risk of accidents. The estimated model explains approximately 94% of the variation in accident counts (R2 = 0.94). The elasticity of passenger car ownership is positive (β = 1.39), indicating increased accident exposure with rising motorization, while paved road density (β = −46.56) and expressways (β = −2.03) show negative elasticities. Average wages are also negatively associated with accidents (β = −4.64). These results quantify the proportional structure of long-term accident dynamics rather than merely confirming directional relationships. The analysis also revealed a negative correlation between urban population density and the number of accidents, which may indicate greater effectiveness of traffic management and control systems in urban areas. The results of the study provide empirical evidence relevant for the development of investment and regulatory strategies in the area of transport infrastructure and road safety.

1. Introduction

Road safety is one of the key elements of modern transport systems. Increased mobility of societies, the development of motorization, and intensified road traffic lead to an increased risk of accidents, which entail significant social, economic, and health costs. According to European data, accident rates remain at a level that requires systemic preventive measures, despite technological progress and infrastructure development. Road accidents are influenced by a complex interplay of infrastructural, economic, and social factors. Quantitative analyses are essential to isolate the impact of each variable [1].
Traffic accidents result in increased healthcare costs, higher demands for vehicle and infrastructure maintenance, and adverse environmental impacts. Therefore, minimizing traffic accidents is a matter of significant concern. One strategy for achieving this goal involves analyzing the factors that contribute to the occurrence of traffic collisions [2,3,4,5].
Research conducted by Zhai et al. [6] and Holland et al. [7] indicates that pedestrians are particularly susceptible compared to vehicle occupants, making them the group at highest risk in traffic incidents. They are also more likely to suffer severe injuries relative to other road users. Various determinants influence the severity of pedestrian injuries, including alcohol consumption, driver age and gender, lighting conditions, road surface quality, pedestrian behavior, accident location, vehicle features, speed, and unfavorable weather conditions [8,9].
Adverse weather and low visibility, especially in the vicinity of pedestrian crossings, are frequently associated with more serious accidents [9,10,11]. Nonetheless, the strength of this relationship may differ depending on the study context. Some studies, for instance [12], indicate that meteorological factors exert only a limited influence on traffic accident rates. Similar analyses by Masello et al. [13] and Becker et al. [14] developed predictive models of accident likelihood based on time of day and weather conditions. Numerous other studies have also explored the link between weather and traffic accidents [15,16,17,18,19,20,21,22,23,24,25,26,27].
While the studies cited above provide valuable insights, most focus on specific determinants such as weather conditions, driver behavior, or pedestrian risk using localized or short-term datasets. Methodologically, many rely on count data models (Poisson or negative binomial), machine learning approaches, or spatial econometric techniques. However, fewer studies examine macro-level, long-term infrastructure expansion jointly with socioeconomic transformation in a unified elasticity-based framework. Moreover, limited attention has been given to the structural co-movement of infrastructure growth and accident decline at the national time-series level. This gap motivates the present analysis.
Moreover, current weather conditions not only affect the frequency of accidents but also influence traffic flow and driver behavior, including reaction times to changing road circumstances [18,28,29]. In a U.S.-based analysis, Eisenberg [20] demonstrated that poor weather conditions are associated with an increase in traffic accident occurrences.
Furthermore, the literature on the subject emphasizes the multifactorial nature of road accidents. Factors determining the number of accidents include both infrastructural elements, such as the quality and density of the road network, the development of expressways, and urbanization conditions, as well as economic and social factors, including income levels, motorization of the population, changes in spatial development, and the behavior of road users. Research shows that changes in the layout and parameters of road infrastructure can lead to both a reduction in the number of accidents (e.g., through modernization and improvement of the transport system) and an increase in accidents if they result in higher average speeds or traffic intensity.
In this context, quantitative models are particularly useful, as they allow for the identification of relationships between the number of accidents and variables representing infrastructure and economic factors. Log–log models based on logarithmic transformation also allow for the direct interpretation of regression coefficients as elasticities, which makes it possible to assess how a percentage change in a given variable affects the percentage change in the number of accidents.
The aim of this article is to empirically assess the impact of selected infrastructural and socio-economic determinants on the number of road accidents in Poland using a logarithmic regression model. The analysis covers variables describing the density and standard of the road network, the level of motorization of the population, and the economic situation of residents. The results obtained allow us to determine the direction and strength of the impact of individual factors and to formulate conclusions of an applied nature, which are important for shaping transport policy and measures to improve road safety.
Rather than testing whether infrastructure and socioeconomic factors influence road accidents—an issue already well established in the literature—this study aims to quantify the relative magnitude, direction, and elasticity of these relationships in a specific national context undergoing dynamic infrastructure expansion and economic transformation. The core objective is therefore not to confirm the existence of effects, but to measure their proportional strength, assess their stability over time, and examine how infrastructure and economic growth jointly shape accident dynamics.
The study focuses on estimating long-run elasticities and evaluating whether infrastructure expansion offsets the risk associated with increasing motorization. This quantitative perspective provides information that cannot be inferred from qualitative reasoning alone.
The remainder of this paper is structured as follows. Section 2 presents the data sources, descriptive statistics, model specification, and diagnostic procedures. Section 3 reports the estimation results, elasticity interpretation, and robustness analysis. Section 4 discusses the implications and methodological considerations. Section 5 concludes with key findings, limitations, and directions for future research.
The contribution of this study lies not in confirming the direction of influence of commonly recognized determinants, but in providing a quantified elasticity-based assessment of their relative magnitude within a unified national time-series framework. Unlike many localized or short-term predictive studies, this approach allows for evaluating proportional long-run relationships between infrastructure growth, socioeconomic transformation, and accident dynamics at the macro level.
These findings clearly indicate how each factor affects accident numbers, providing actionable guidance for transport policy.

Scientific Contribution

The contribution of this study does not lie in demonstrating that infrastructure and economic factors influence road safety, which is well recognized in the literature. Instead, the novelty consists of three elements.
First, the study provides quantitative elasticity estimates for infrastructure and socioeconomic variables within a unified macro-level framework. While individual determinants have been widely studied, fewer analyses estimate their proportional effects simultaneously using consistent time-series data.
Second, the analysis captures a period characterized by rapid expressway expansion, sustained economic growth, and a structural shock caused by the COVID-19 pandemic. This creates an opportunity to examine how infrastructure development interacts with macroeconomic conditions under non-stationary dynamics.
Third, the study highlights the presence of strong co-movements and multicollinearity among infrastructure and economic indicators, demonstrating that macro-level accident reductions are embedded in broader structural transformations rather than isolated interventions. This methodological insight contributes to the discussion on how national-level accident models should be specified and interpreted.
Thus, the study advances the field not by confirming well-known relationships, but by quantifying their magnitude, examining their joint evolution, and critically evaluating their statistical structure.

2. Materials and Methods

The study analyzes the relationship between road traffic accidents and selected infrastructural and economic factors in Poland for the period 2010–2024 (Table 1). The dependent variable is the annual total number of reported road accidents (WYP). The independent variables include:
  • − DRT—paved roads per 100 km2,
  • − EX—expressways and motorways per 100 km2,
  • − DEN—density of built-up and urbanized areas (persons/km2),
  • − CARS—passenger cars per 1000 inhabitants,
  • − WAGE—average gross wage (PLN).
Data were collected from official statistical sources, including the Polish Central Statistical Office (GUS) and national road infrastructure databases. The dataset covers a 15-year period from 2010 to 2024, capturing trends in road infrastructure, vehicle ownership, population distribution, and economic conditions.
Table 1. Statistical data from 2010 to 2024 [1,2,3,30].
It should be noted that DEN represents the ratio of built-up and urbanized land to total area rather than population density per square kilometer. The values reported in Table 1 therefore reflect proportional land-use intensity rather than demographic density.
All values reported in Table 1 represent annual national-level observations and are not computed as averages across regions. The data correspond to officially reported yearly totals or national aggregates.

2.1. Descriptive Statistics and Trends

The dataset used in this study covers the period from 2010 to 2024, and reflects substantial changes in road safety, infrastructure development, and economic conditions in Poland. Over this time span, the total number of reported road accidents shows a clear downward trend, decreasing from nearly 39,000 accidents in 2010 to around 21,500 in 2024. Although some year-to-year fluctuations are visible, the overall pattern indicates a long-term improvement in road safety at the national level.
A particularly noticeable decline in the number of accidents occurred in 2020 and 2021. This period coincides with the COVID-19 pandemic, when mobility restrictions, remote work, and reduced traffic volumes significantly limited road use. While accident numbers slightly increased again in subsequent years, they did not return to pre-pandemic levels, suggesting that part of the observed reduction may be associated with longer-lasting changes in travel behavior.
At the same time, the descriptive data show a steady expansion of road infrastructure. The density of paved roads increased gradually over the analyzed period, reflecting continuous investments in the road network. An even more pronounced growth can be observed in the length of expressways and motorways per 100 km2, which more than tripled between 2010 and 2024. This dynamic development illustrates a structural transformation of the Polish road system toward higher technical standards.
Urban population density shows a mild but consistent decline over the period under study. This trend reflects ongoing processes of suburbanization and spatial dispersion of the population rather than rapid urban growth. Despite this decrease in density, road accident numbers continued to fall, indicating that factors other than population concentration play an important role in shaping road safety outcomes.
Vehicle ownership increased steadily throughout most of the analyzed period, with the number of passenger cars per 1000 inhabitants rising significantly between 2010 and 2023. This trend reflects growing motorization and increased mobility. A temporary decline observed in 2024 may be linked to data revisions or short-term economic factors. Despite the increase in vehicle ownership, the overall number of accidents decreased, suggesting that improvements in infrastructure and economic conditions may have partly offset the safety risks associated with higher traffic exposure.
Economic conditions improved markedly during the study period, as reflected in the strong growth of average gross wages. Rising income levels may be associated with safer vehicles, better maintenance, and improved compliance with traffic regulations. Taken together, the descriptive statistics highlight that the analyzed period was characterized by increasing motorization and economic growth alongside a substantial reduction in road accidents, providing a strong motivation for the subsequent econometric analysis.

2.2. Model Specification

To examine the impact of infrastructural and economic variables on road accident numbers, a log–linear regression model was specified as follows:
l n ( W Y P t ) = β 0 + β 1 l n ( D R T t ) + β 2 l n ( E X t ) + β 3 l n ( D E N t ) + β 4 l n ( C A R S t ) + β 5 l n ( W A G E t )
where l n ( W Y P t ) —natural logarithm of the number of traffic accidents in year t; D R T t , E X t , D E N t ,   C A R S t ,   W A G E t —independent variables as defined above; and β 0 , β 1 , … , β 5 —regression coefficients to be estimated.
Equation (1) presents the theoretical log–log specification of the model.
Alternative specifications, including semi-log and linear models, were initially tested. However, these specifications resulted in lower explanatory power and less stable residual structures compared to the log–log model. Therefore, the elasticity-based log–log formulation was retained as the primary specification.
Although count-data models such as Poisson or negative binomial regression are commonly used in accident analysis, the log–log specification was selected to allow direct elasticity interpretation within a macro-level framework.

2.3. Econometric Assumptions and Model Diagnostics

The econometric analysis is based on a log–log regression model estimated using the Ordinary Least Squares (OLS) method. This specification was chosen because it allows for a flexible representation of potentially nonlinear relationships between road accidents and their determinants, while at the same time providing coefficients that can be interpreted as elasticities. Such an approach is well suited for analyzing macro-level relationships using annual data.
To assess the potential presence of multicollinearity among explanatory variables, a Pearson correlation matrix was computed for the logarithmically transformed variables. The results indicate strong pairwise correlations between several infrastructure and economic variables, reflecting their common upward time trend over the analyzed period. In particular, paved road density (DRT), expressways and motorways (EX), vehicle ownership (CARS), and wages (WAGE) exhibit high positive correlations, suggesting that infrastructure development and economic growth evolved jointly over time.
Variance Inflation Factors (VIFs) were calculated to assess multicollinearity among the explanatory variables. The results, presented in Table 2, indicate substantial collinearity, particularly for WAGE (22.6), DRT (18.7), and DEN (16.9), exceeding the commonly used threshold of 10. These findings confirm that infrastructure and economic variables share strong common time trends, which may inflate coefficient magnitudes in the baseline specification.
Table 2. Variance Inflation Factors (VIFs).
Given the limited number of observations (15 annual data points) and the strong time trend in several variables, part of the large elasticity estimates may reflect common long-term dynamics rather than purely causal effects. Therefore, additional robustness checks were conducted to assess the stability of the coefficients.
Prior to estimation, all variables were transformed using natural logarithms. This transformation helps to stabilize variance, reduce the influence of extreme values, and linearize multiplicative relationships between the dependent and independent variables. As a result, the estimated model focuses on proportional rather than absolute changes, which is particularly appropriate when analyzing long-term trends in infrastructure, motorization, and economic development.
Standard diagnostic checks were conducted to assess the validity of the OLS assumptions. In particular, the potential presence of multicollinearity among explanatory variables was examined, given that infrastructure development, motorization, and income tend to evolve jointly over time. While standard residual diagnostics (e.g., heteroskedasticity tests and specification checks) did not reveal serious violations of classical OLS assumptions, the VIF results confirm the presence of substantial multicollinearity among selected regressors. Therefore, the primary concern relates to the precision and magnitude of coefficient estimates rather than the direction of their effects.
In addition, the model was evaluated for potential heteroskedasticity and overall specification adequacy. Given the use of aggregated annual data and logarithmic transformation, the residuals did not exhibit patterns that would suggest severe violations of the classical regression assumptions. The estimated model therefore provides a reasonable approximation of the underlying relationships between the analyzed variables and the number of road accidents.
It should be noted that the dependent variable represents annual counts of road accidents, which are inherently discrete in nature. While the log–log OLS framework is commonly applied in macro-level road safety studies and allows for straightforward elasticity interpretation, alternative modeling approaches—such as Poisson or negative binomial regression—may be considered in future research to explicitly account for the count-data characteristics and potential overdispersion.
Overall, despite the presence of multicollinearity typical for trending macro-level series, the diagnostic assessment supports the use of the specified log–linear model as a transparent framework for examining long-term proportional relationships.
The goodness-of-fit statistics show how well the log–log regression model captures the observed variation in the number of road accidents (Table 3). The R2 value indicates that approximately 94% of the variation in the logarithm of accident counts is explained by the model, reflecting a strong fit. The adjusted R2 accounts for the number of predictors and confirms that the model’s explanatory power remains high even after considering model complexity. The F-statistic tests whether all regression coefficients are jointly significant, and the very low p-value (<0.001) demonstrates that the model as a whole is statistically meaningful. Finally, the standard error of the regression measures the typical deviation of observed log-accident values from the model’s predicted values, providing an indication of the accuracy of the model’s predictions.
Table 3. Model Goodness-of-Fit Statistics.
Overall, these statistics suggest that the selected infrastructure and economic variables collectively explain most of the long-term trends in road accidents in Poland over the 2010–2024 period, reinforcing the relevance of including both infrastructural and socio-economic factors in road safety analyses. Given the relatively small sample size and the trending nature of macro-level variables, the presence of multicollinearity is not unexpected. Therefore, robustness tests were conducted to verify the qualitative stability of coefficient signs.

3. Results

The estimated log–linear regression model for the number of road accidents in Poland is as follows:
l n ( W Y P t ) = 580.115 − 46.5565 l n ( D R T t ) − 2.032 l n ( E X t ) − 42.2108 l n ( D E N t ) + 1.3882 l n ( C A R S t ) − 4.6409 l n ( W A G E t )
The unusually high magnitude of the DRT coefficient may reflect the logarithmic transformation of variables and the scale of paved road data.
Equation (2) represents the empirical OLS estimates of the theoretical model specified in Equation (1). Each coefficient corresponds to the estimated partial elasticity of accident counts with respect to the respective explanatory variable, conditional on other variables included in the model. All coefficients reported in Equation (2) were obtained using Ordinary Least Squares (OLS) estimation applied to natural logarithms of the variables. The estimation was conducted on annual national-level data for the period 2010–2024. No additional transformations, weighting procedures, or regularization techniques were applied. Standard errors were computed under classical OLS assumptions.
Based on the above formula, the following conclusions were drawn (Table 4):
-
DRT (paved roads per 100 km2, β1 = −46.5565): The estimated coefficient suggests a very high negative elasticity. In a log–log specification, this value formally implies that a 1% increase in paved road density is associated with a 46% decrease in accident counts. However, given the strong upward time trend in infrastructure variables and the relatively small variation in DRT over the analyzed period, such a magnitude should be interpreted as reflecting the dominant long-term co-movement between infrastructure expansion and declining accident numbers rather than a short-run causal elasticity. The unusually large coefficient indicates potential multicollinearity and trend-driven effects in the model.
-
EX (expressways and motorways per 100 km2, β2 = −2.0320): An increase of 1% in the length of expressways and motorways is associated with a 2.03% reduction in accidents. While the effect is smaller than for general paved roads, it indicates that higher-standard roads contribute to improved safety.
-
DEN (density of built-up areas, β3 = −42.2108): The estimated coefficient indicates a large negative elasticity. However, similarly to DRT, the magnitude of this parameter likely reflects common long-term structural changes rather than a direct proportional short-term effect. Urban density evolved gradually over time, and its strong correlation with other macro-level variables may inflate the estimated elasticity. Therefore, the result should be interpreted as evidence of a negative association rather than a precise quantitative elasticity.
-
CARS (passenger cars per 1000 inhabitants, β4 = 1.3882): A 1% increase in the number of cars per 1000 inhabitants leads to a 1.39% increase in accidents, indicating that higher vehicle ownership raises exposure to traffic risks.
-
WAGE (average gross wage, β5 = −4.6409): A 1% increase in average gross wage is associated with a 4.64% decrease in accidents, suggesting that higher income levels may be linked to safer driving conditions, better vehicle quality, or increased investment in road safety.
Table 4. Results of the logarithmic regression model for the number of road accidents in Poland (2010–2024).
We interpret the coefficients in the logarithmic model as elasticities: the percentage change in the number of accidents in response to a 1% change in a given independent variable. All variables are statistically significant at the 5% significance level (p < 0.05).
The results indicate that both infrastructural and economic variables significantly influence road accident numbers in Poland. Improvements in road infrastructure (DRT, EX) and higher economic status (WAGE) are associated with lower accident rates, whereas increased vehicle ownership (CARS) is linked to higher accident risk. The negative coefficient for DEN highlights potential safety advantages of urbanized, densely managed areas. These findings provide empirical support for policy measures focusing on road infrastructure development, traffic management in urban areas, and economic incentives to improve road safety.

3.1. Elasticity Interpretation and Sensitivity

The coefficients estimated in the log–log regression model can be interpreted directly as elasticities, which allows us to understand how proportional changes in infrastructure or economic variables are associated with proportional changes in the number of road accidents. This provides a clear and intuitive measure of the sensitivity of accident counts to each factor included in the model.
For example, DRT shows a strong negative elasticity, highlighting that improvements in general road quality substantially reduce accident risk. While the numerical magnitude may appear large, it reflects the strong long-term influence of general road quality on traffic safety, highlighting the substantial role of infrastructure in reducing accident risk.
The coefficient for expressways and motorways (EX) is also negative, though smaller in magnitude. A 1% increase in high-standard roads is associated with about a 2% reduction in accidents. This suggests that, while the expansion of expressways contributes to safety, its effect is less pronounced than improvements in overall paved road density, likely because accidents on expressways constitute a smaller share of total incidents.
Urban population density (DEN) shows a negative elasticity as well, which may seem counterintuitive. This result indicates that denser urban areas experience fewer accidents, possibly due to lower vehicle speeds, better traffic management, and enhanced pedestrian infrastructure. It emphasizes the potential safety benefits of well-planned, compact urban environments.
In contrast, vehicle ownership (CARS) has a positive coefficient, meaning that increased motorization leads to higher accident exposure. A 1% rise in the number of passenger cars per 1000 inhabitants corresponds to a roughly 1.4% increase in accidents, underscoring that higher traffic density elevates risk even when infrastructure improves.
Finally, higher average wages (WAGE) are associated with fewer accidents. The negative elasticity suggests that better economic conditions may promote safer vehicles, improved maintenance, and more responsible driving behavior, all of which contribute to lower accident numbers.
Overall, interpreting the coefficients as elasticities allows policymakers to understand not only the direction but also the relative magnitude of the effects of infrastructure, urbanization, motorization, and economic conditions on road safety. This elasticity-based perspective also provides a useful tool for assessing which interventions are likely to have the greatest impact and helps guide investment priorities in road infrastructure and safety programs. It should be emphasized that the large absolute values of the DRT and DEN elasticities should not be interpreted as short-run causal proportional effects. These coefficients reflect long-term structural co-movement between infrastructure expansion and accident reduction within a trending macro time-series framework. Given the relatively small sample size (n = 15) and high correlation among explanatory variables, the magnitude of these elasticities may partially capture overlapping temporal dynamics rather than isolated marginal effects.

3.2. Robustness and Scenario Analysis

To assess the robustness of the estimated elasticities, additional model specifications were estimated by excluding selected years characterized by structural disturbances or atypical observations.
First, the year 2020 was excluded due to the sharp decline in mobility caused by COVID-19 restrictions. Second, both 2020 and 2021 were removed simultaneously to account for prolonged pandemic effects. Third, the year 2024 was excluded due to the observed decline in vehicle ownership (CARS), which may represent a structural break or reporting adjustment. Finally, a restricted specification excluding 2020, 2021, and 2024 was estimated to test maximum sensitivity.
Table 4 presents a comparison of coefficient estimates, goodness-of-fit statistics, and percentage deviations from the baseline model (2010–2024).
The results indicate that the signs of all coefficients remain unchanged across specifications, suggesting qualitative stability of the relationships. However, the magnitude of infrastructure-related elasticities decreases when pandemic years are excluded, indicating that part of the strong baseline elasticities is driven by the sharp accident reduction observed during COVID-19.
Changes in coefficient magnitude were evaluated using relative percentage differences compared to the baseline model. Elasticity shifts exceeding 30% are interpreted as evidence of parameter sensitivity. The analysis shows that infrastructure coefficients exhibit moderate instability under sample restrictions, whereas the elasticity of motorization (CARS) remains comparatively stable (Table 5).
Table 5. Robustness analysis—comparison of alternative sample specifications.
Overall, while the direction of effects proves robust, the magnitude of long-run elasticities is sensitive to structural breaks and small-sample dynamics. This finding highlights the importance of cautious interpretation in macro-level time-series models with limited observations.
Importantly, although the magnitude of infrastructure elasticities decreases after excluding pandemic years, the direction of influence remains unchanged in all specifications. This suggests qualitative robustness of the structural relationships despite quantitative sensitivity in small samples. The maximum absolute percentage deviation (|Δβ|) was calculated relative to the baseline specification. Infrastructure-related elasticities (DRT and DEN) show changes exceeding 40% after excluding pandemic years, indicating quantitative sensitivity in a small-sample macro time-series setting. However, coefficient signs remain unchanged and all variables retain statistical significance at the 5% level, confirming qualitative robustness of the model.

3.3. Estimation Procedure

The coefficients reported in Equation (2) were obtained by estimating the log–log specification using Ordinary Least Squares (OLS) applied to annual data for the period 2010–2024 (n = 15). All variables were transformed using natural logarithms prior to estimation. The estimation was performed using standard statistical software 13.3, and coefficient values correspond to the OLS estimators that minimize the sum of squared residuals.
Standard errors, t-statistics, and p-values were calculated under classical OLS assumptions. Model diagnostics, including multicollinearity assessment (VIF), residual analysis, and goodness-of-fit statistics, are reported in subsequent sections.
The log-transformation of variables allows for interpreting the estimated coefficients as elasticities, i.e., the percentage change in the number of accidents associated with a 1% change in each independent variable.
The model parameters were estimated using the Ordinary Least Squares (OLS) method. Prior to estimation, all variables were transformed using natural logarithms to linearize potential multiplicative relationships and stabilize variance. Diagnostic checks were conducted to verify the assumptions of OLS regression, including multicollinearity, heteroscedasticity, and model specification.

4. Discussion

The analysis of the log–linear regression model reveals several important insights regarding the relationship between infrastructural, economic factors, and road accident numbers in Poland over the period 2010–2024.
It is important to emphasize that the quantitative results obtained in this study are not self-evident. While it is intuitively plausible that infrastructure expansion reduces accidents and motorization increases exposure, the magnitude, proportional structure, and joint dynamics of these effects cannot be deduced without empirical estimation. In particular, the analysis reveals:
-
The dominance of long-term infrastructure trends in explaining accident decline;
-
The relative elasticity differences between general paved roads and expressways;
-
The interaction between economic growth and accident reduction;
-
The statistical instability introduced by strong time co-movements.
These insights emerge only through formal modeling and cannot be inferred from qualitative reasoning alone.
The negative coefficient for DRT (paved roads per 100 km2) indicates that an increase in the length of paved roads is strongly associated with a reduction in traffic accidents. This finding aligns with previous research suggesting that high-quality road infrastructure facilitates safer driving conditions by reducing hazards, improving vehicle handling, and allowing smoother traffic flow (e.g., [31]). The estimated effect size, approximately 46.6% per 1% increase in logarithmic terms, is substantial; however, such high elasticity values reflect the log-transformed nature of the variables and should be interpreted as indicative of the strong negative relationship rather than a literal 46% decrease per 1% increase in road length.
Similarly, the EX (expressways and motorways per 100 km2) variable shows a negative, albeit smaller, effect on accident numbers. This suggests that high-standard roads designed for faster traffic with controlled access and fewer conflict points contribute to reduced accident risk. Although the effect size is modest (−2.03% per 1% increase), it supports the notion that investments in expressways and motorways are beneficial for road safety.
Interestingly, DEN (share of built-up and urbanized areas in total land area (expressed as a ratio)) has a strong negative coefficient, although higher urban density might be expected to increase accidents due to more vehicles and interactions. This suggests that denser urban areas may benefit from effective traffic management, lower vehicle speeds, better pedestrian infrastructure, and enhanced surveillance, collectively reducing accident frequency. However, the results suggest that denser urbanized areas may benefit from traffic management measures, lower speeds, better surveillance, and more pedestrian infrastructure, which collectively reduce accident severity and frequency. This finding is consistent with studies highlighting the positive role of urban planning in mitigating traffic risk.
The CARS (cars per 1000 inhabitants) coefficient is positive, indicating that increased vehicle ownership contributes to higher accident exposure, as expected. More cars on the road increase the likelihood of interactions and collisions. The relatively small elasticity (1.39% per 1% increase) suggests that while higher vehicle density raises risk, other factors such as infrastructure quality and traffic regulations may moderate this effect.
Higher income appears to improve road safety, likely through safer vehicles and better driving conditions. Higher wages may enable households to purchase safer, modern vehicles, invest in driver education, and support better traffic infrastructure through taxation and private investment. This aligns with previous findings that socioeconomic development is associated with improved road safety outcomes.
The results underline the importance of integrating infrastructural and economic considerations in road safety policies. Investments in paved roads and expressways, along with urban planning measures that manage traffic flow in densely built areas, can substantially reduce accident rates. Furthermore, socioeconomic improvements, such as higher wages and overall prosperity, indirectly contribute to safety by enhancing vehicle quality and enabling better adherence to traffic rules.
Despite the robustness of the model, several limitations should be noted. First, the analysis relies on aggregate annual data, which may mask regional variations in accident patterns. Future studies could employ disaggregated, city- or province-level data to capture local effects. Second, other potential determinants, such as driver behavior, alcohol consumption, vehicle age, and weather conditions, were not included due to data limitations. Incorporating these factors could provide a more comprehensive understanding of accident causation. Finally, while the log–linear specification allows interpretation of elasticities, alternative models (e.g., Poisson or negative binomial regression) could be explored to account for the discrete nature of accident counts.

Policy Implications

The findings of this study offer clear guidance for policymakers seeking to improve road safety in Poland. Firstly, investments in general road infrastructure, including the expansion and maintenance of paved roads, have a strong and direct effect on reducing accident numbers. Prioritizing projects that improve the quality and coverage of the road network can therefore yield substantial safety benefits.
Secondly, although expressways and motorways contribute to safety, their relative impact is smaller than that of overall paved roads. This suggests that while high-standard roads are important, broader improvements across the entire network are essential for achieving nationwide reductions in accidents.
Thirdly, the negative relationship between urban population density and accidents highlights the potential benefits of compact, well-managed urban areas. Urban planning measures that reduce vehicle speeds, improve pedestrian and cyclist infrastructure, and enhance traffic management can effectively reduce accidents even in densely populated cities. Policymakers should therefore integrate traffic safety considerations into urban development strategies.
Fourthly, the positive effect of vehicle ownership on accident numbers underscores the need for complementary measures to manage motorization growth. Policies such as promoting public transport, encouraging safer vehicle technologies, and implementing traffic-calming measures can mitigate the risks associated with increased car use.
Finally, higher economic status, as reflected in average wages, is associated with fewer accidents. This indicates that socio-economic development itself can contribute to road safety, possibly by enabling households to afford safer vehicles, better maintenance, and driver education. Policymakers should consider strategies that combine infrastructure investment with broader social and economic initiatives to maximize road safety improvements.
In summary, the study suggests that effective road safety policy should adopt a holistic approach, combining infrastructure development, urban planning, traffic management, and socio-economic measures. Such integrated strategies are likely to produce the largest and most sustainable reductions in traffic accidents over time.
While the direction of influence of infrastructure and motorization on accidents is widely acknowledged, the magnitude of elasticities observed in this study provides policymakers with quantitative benchmarks that can inform proportional infrastructure investment strategies and long-term safety planning.

5. Conclusions

The study examined the impact of infrastructural and economic factors on road accident numbers in Poland for the period 2010–2024 using a log–linear regression model. The analysis focused on the effects of paved roads (DRT), expressways and motorways (EX), urban density (DEN), vehicle ownership (CARS), and average gross wages (WAGE) on annual accident counts.
Future research could utilize regional or city-level data to capture local variations in accident patterns, as aggregate annual data may mask important local effects. Incorporating additional variables, such as weather conditions, driver behavior (including alcohol consumption and adherence to traffic rules), vehicle characteristics, traffic volume, and land use, would provide a more comprehensive understanding of accident determinants.
Alternative modeling approaches, including Poisson or negative binomial regressions, could better account for the discrete and potentially over-dispersed nature of accident counts, while advanced machine learning methods might reveal nonlinear relationships and interactions among factors, supporting more targeted road safety policies.
Instead of testing obvious directional relationships, the study provides quantitative evidence on the relative strength and structural interdependence of infrastructure and economic variables in shaping national accident trends. The findings demonstrate that accident reductions in Poland during 2010–2024 cannot be attributed to a single factor but reflect a broader structural transformation involving infrastructure expansion, economic growth, and evolving traffic exposure.
The study also highlights methodological challenges associated with macro-level accident modeling, particularly multicollinearity and trend-driven elasticities, contributing to a more cautious interpretation of national time-series models.

Limitations and Future Research

While the present study provides important insights into the relationship between infrastructure, economic conditions, and road accidents, several limitations should be acknowledged.
First, the analysis is based on 15 annual national-level observations (2010–2024), which restricts degrees of freedom and increases sensitivity to structural breaks and common time trends. With multiple explanatory variables evolving jointly over time, the risk of multicollinearity and inflated elasticity estimates cannot be excluded.
Second, the model relies on aggregate macro-level indicators and does not include several potentially important determinants of accident occurrence, such as traffic volume, enforcement intensity, driver behavior, alcohol consumption, vehicle age structure, or detailed weather conditions. The absence of these variables reflects data constraints and may limit the causal interpretation of coefficients.
Third, although the log–log OLS framework is suitable for analyzing long-term proportional relationships, accident counts are discrete and may exhibit overdispersion. Alternative approaches, such as Poisson or negative binomial regression, panel-data models, or nonlinear machine learning techniques, could better capture distributional properties and interaction effects.
Future research should therefore extend the analysis using longer time series, regional panel data, or cross-country comparisons to improve robustness and external validity. Scenario-based simulations combining infrastructure and socio-economic interventions could further support strategic road safety planning.
The results are particularly useful for national policymakers, transport planners, and infrastructure investment authorities seeking to assess the proportional long-term impact of infrastructural expansion and economic development on road safety outcomes.

Author Contributions

Conceptualization, P.G. and E.S.; methodology, P.G. and E.S.; software, P.G.; validation, P.G. and E.S.; formal analysis, P.G. and E.S.; investigation, P.G. and E.S.; resources, P.G. and E.S.; data curation, P.G. and E.S.; writing—original draft preparation, P.G. and E.S.; writing—review and editing, P.G. and E.S.; visualization, P.G. and E.S.; supervision, P.G. and E.S.; project administration, P.G. and E.S.; funding acquisition, P.G. and E.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

All data used in this study are publicly available from the Polish Central Statistical Office (GUS), the Police statistical database, and Eurostat. The compiled dataset covering the period 2010–2024 is available from the authors upon reasonable request.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

DRTDensity of Paved Roads
EXExpressways and Motorways
DENBuilt-up Area Ratio
CARSPassenger Cars per 1000 Inhabitants
WAGEAverage Gross Wage
WYPRoad Accidents
OLSOrdinary Least Squares
VIFVariance Inflation Factor

References

  1. Sewik. Available online: https://sewik.pl (accessed on 6 January 2025).
  2. Police Statistics. Available online: https://statystyka.policja.pl/ (accessed on 6 January 2025).
  3. Eurostat. Available online: https://ec.europa.eu/eurostat (accessed on 6 January 2025).
  4. Gorzelańczyk, P.; Sokolovskij, E. Using neural networks to forecast the amount of traffic accidents in Poland and Lithuania. Sustainability 2025, 17, 1846. [Google Scholar] [CrossRef] [Scilit]
  5. Melders, L.; Smigins, R.; Birkavs, A. Recent advances in vehicle driver health monitoring systems. Sensors 2025, 25, 1812. [Google Scholar] [CrossRef] [Scilit]
  6. Zang, S.; Ding, M.; Smith, D.; Tyler, P.; Rakotoarivelo, T.; Kaafar, M.A. The impact of adverse weather conditions on autonomous vehicles: How rain, snow, fog, and hail affect the performance of a self-driving car. IEEE Veh. Technol. Mag. 2019, 14, 103–111. [Google Scholar] [CrossRef] [Scilit]
  7. Zhai, X.; Huang, H.; Sze, N.N.; Song, Z.; Hon, K.K. Diagnostic analysis of the effects of weather condition on pedestrian crash severity. Accid. Anal. Prev. 2019, 122, 318–324. [Google Scholar] [CrossRef] [Scilit]
  8. Holland, C.; Hill, R. The effect of age, gender and driver status on pedestrians’ intentions to cross the road in risky situations. Accid. Anal. Prev. 2007, 39, 224–237. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  9. Amini, R.E.; Katrakazas, C.; Antoniou, C. Negotiation and decision-making for a pedestrian roadway crossing: A literature review. Sustainability 2019, 11, 6713. [Google Scholar] [CrossRef] [Scilit]
  10. Favarò, F.M.; Nader, N.; Eurich, S.O.; Tripp, M.; Varadaraju, N. Examining accident reports involving autonomous vehicles in California. PLoS ONE 2017, 12, e0184952. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  11. Hafeez, F.; Sheikh, U.U.; Al-Shammari, S.; Hamid, M.; Khakwani, A.B.K.; Arfeen, Z.A. Comparative analysis of influencing factors on pedestrian road accidents. Bull. Electr. Eng. Inform. 2023, 12, 257–267. [Google Scholar] [CrossRef] [Scilit]
  12. Mesquitela, J.; Elvas, L.B.; Ferreira, J.C.; Nunes, L. Data analytics process over road accidents data—A case study of Lisbon city. ISPRS Int. J. Geo-Inf. 2022, 11, 143. [Google Scholar] [CrossRef] [Scilit]
  13. Masello, L.; Castignani, G.; Sheehan, B.; Murphy, F.; McDonnell, K.L. On the road safety benefits of advanced driver assistance systems in different driving contexts. Transp. Res. Interdiscip. Perspect. 2022, 15, 100670. [Google Scholar] [CrossRef] [Scilit]
  14. Becker, N.; Rust, H.W.; Ulbrich, U. Predictive modeling of hourly probabilities for weather-related road accidents. Nat. Hazards Earth Syst. Sci. 2020, 20, 2857–2871. [Google Scholar] [CrossRef] [Scilit]
  15. Mills, B.; Andrey, J.; Doberstein, B.; Doherty, S.; Yessis, J. Changing patterns of motor vehicle collision risk during winter storms: A new look at a pervasive problem. Accid. Anal. Prev. 2019, 127, 186–197. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  16. Karlaftis, M.; Yannis, G. Weather effects on daily traffic accidents and fatalities: A time series count data approach. In Proceedings of the 89th Annual Meeting of the Transportation Research Board, Washington, DC, USA, 10–14 January 2010. [Google Scholar]
  17. Scott, P. Modelling time-series of British road accident data. Accid. Anal. Prev. 1986, 18, 109–117. [Google Scholar] [CrossRef] [Scilit]
  18. Fridstrøm, L.; Ifver, J.; Ingebrigtsen, S.; Kulmala, R.; Thomsen, L.K. Measuring the contribution of randomness, exposure, weather, and daylight to the variation in road accident counts. Accid. Anal. Prev. 1995, 27, 1–20. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  19. Shankar, V.; Mannering, F.; Barfield, W. Effect of roadway geometrics and environmental factors on rural freeway accident frequency. Accid. Anal. Prev. 1995, 27, 371–389. [Google Scholar] [CrossRef] [Scilit]
  20. Eisenberg, D. The mixed effects of precipitation on traffic accidents. Accid. Anal. Prev. 2004, 36, 637–647. [Google Scholar] [CrossRef] [Scilit]
  21. Bergel-Hayat, R.; Depire, A. Climate, road traffic and road risk—An aggregate approach. In Proceedings of the 10th World Conference on Transport Research, 2004 (CD-ROM), Istanbul, Turkey, 4–6 July 2004. [Google Scholar]
  22. Hermans, E.; Brijs, T.; Stiers, T.; Offermans, C. The impact of weather conditions on road safety investigated on an hourly basis. In Proceedings of the 85th Transportation Research Board (TRB) Annual Meeting, Washington, DC, USA, 22–26 January 2006. [Google Scholar]
  23. Bijleveld, F.; Churchill, T. The Influence of Weather Conditions on Road Safety: An assessment of the Effect of Precipitation and Temperature. 2009; p. 47. Available online: https://swov.nl/system/files/publication-downloads/R-2009-09.pdf (accessed on 6 January 2025).
  24. Bucsuházy, K.; Matuchová, E.; Zuvala, R.; Moravcová, P.; Kostíková, M.; Mikulec, R. Human factors contributing to the road traffic accident occurrence. Transp. Res. Procedia 2020, 45, 555–561. [Google Scholar] [CrossRef] [Scilit]
  25. Schlogl, M. A multivariate analysis of environmental effects on road accident occurrence using a balanced bagging approach. Accid. Anal. Prev. 2020, 136, 105398. [Google Scholar] [CrossRef] [Scilit]
  26. Kim, D.; Jung, S.; Yoon, S. Risk prediction for winter road accidents on expressways. Appl. Sci. 2021, 11, 9534. [Google Scholar] [CrossRef] [Scilit]
  27. Kim, D.; Yoon, S.; Kim, B. Comparison of spatial interpolation methods for producing road weather information in winter. J. Korean Data Anal. Soc. 2020, 23, 541–551. [Google Scholar] [CrossRef] [Scilit]
  28. Hayat, R.; Debbarh, M.; Antoniou, C.; Yannis, G. Explaining the road accident risk: Weather effects. Accid. Anal. Prev. 2013, 60, 456–465. [Google Scholar] [CrossRef] [Scilit]
  29. Hermans, E.; Wets, G.; Van den Bossche, F. Frequency and severity of Belgian road traffic accidents studied by state-space methods. J. Transp. Stat. 2006, 9, 63–76. [Google Scholar]
  30. Polish Central Statistical Office (GUS). Local Data Bank (BDL). Available online: https://bdl.stat.gov.pl (accessed on 6 January 2025).
  31. Huynh, V.; Truong, L.; De Gruyter, C. Examining the impacts of road pavement roughness and rutting on traffic safety: A macrolevel analysis. Traffic Inj. Prev. 2025, 26, 720–726. [Google Scholar] [CrossRef] [Scilit]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.